Endogeneity of acreage choices in input allocation equations: implied problems and a solution

Standard farm-level data used for crop choice modelling are farm accountancy data providing detailed data with respect to production (crop areas and production) but only providing variable input use (pesticides, fertilizers, …) data at the farm level. Allocation of variable input uses to crops appears useful for several reasons: it is necessary to compute gross margins at the crop level and may be used to investigate the empirical validity of multi-crop econometric models. The contribution of this research is threefold. First it shows that the standard regression based approaches for allocating variable input uses to crops are potentially biased due to the (partial) simultaneity of the (expected) crop variable input and acreage choices. Second, it proposes a econometric multi-crop model which is consistent with respect to its deterministic and random parts, for defining these biases and for providing potential solutions. Third, it proposes approaches to eliminate these biases. These approaches are based on the use of correction or control functions (Imbens and Wooldridge 2007; Wooldridge 2008).